bank cycles 2009
TRANSCRIPT
-
7/29/2019 Bank Cycles 2009
1/62
Alistair Milne Geoffrey Wood
The bank lending channel
reconsidered
Bank of Finland ResearchDiscussion Papers2 2009
-
7/29/2019 Bank Cycles 2009
2/62
Suomen PankkiBank of Finland
PO Box 160FI-00101 HELSINKI
Finland
+358 10 8311
http://www.bof.fi
-
7/29/2019 Bank Cycles 2009
3/62
Bank of Finland ResearchDiscussion Papers22009
Alistair Milne* Geoffrey Wood**
The bank lending channel
reconsidered
The views expressed in this paper are those of the authors anddo not necessarily reflect the views of the Bank of Finland.
* Cass Business School, UK and Monetary Policy and
Research Department, Bank of Finland, Helsinki.Corresponding author. E-mail: [email protected].** Cass Business School, UK and Monetary Policy and
Research Department, Bank of Finland, Helsinki.
We are grateful to Josephine Fogden and Tarja Yrjl forexcellent research assistance and to comments from JeremyStein and from the internal workshop of the Bank of Finlandresearch department.
-
7/29/2019 Bank Cycles 2009
4/62
http://www.bof.fi
ISBN 978-952-462-484-8ISSN 0785-3572
(print)
ISBN 978-952-462-485-5ISSN 1456-6184(online)
Helsinki 2009
-
7/29/2019 Bank Cycles 2009
5/62
3
The bank lending channel reconsidered
Bank of Finland ResearchDiscussion Papers 2/2009
Alistair Milne Geoffrey Wood
Monetary Policy and Research Department
Abstract
It has been widely accepted that constraints on the wholesale funding of bankbalance sheets amplify the transmission of monetary policy through what is called
the bank lending channel. We show that the effect of such bank balance sheet
constraints on monetary transmission is in fact theoretically ambiguous, with the
prior expectation, based on standard theoretical models of household and
corporate portfolios, that the bank lending channel attenuates monetary policy
transmission.
We examine macroeconomic data for the G8 countries and find no evidence
that banking sector deposits respond negatively and more than lending to
tightening of monetary policy, as the accepted view of the bank lending channel
requires. The overall picture is mixed, but these data generally suggest that
deposits fluctuate procyclically and somewhat less over the business cycle than
bank lending, and that total bank deposits, unlike bank lending, show little direct
response to changes in interest rates. This suggests it is very unlikely that the bank
lending channel amplifies monetary policy. Our paper has thus corrected a
misunderstanding about the role of banks in monetary policy transmission that has
persisted in the literature for some two decades.
Keywords: credit channel, monetary transmission, bank financing constraints
JEL classification numbers: E44, E52, G32
-
7/29/2019 Bank Cycles 2009
6/62
4
Uudelleenarviointi: pankkien merkitys rahapolitiikanvaikutusten vlittymisess
Suomen Pankin keskustelualoitteita 2/2009
Alistair Milne Geoffrey Wood
Rahapolitiikka- ja tutkimusosasto
Tiivistelm
Suhteellisen yleisen ksityksen mukaan pankkeihin kohdistuvat tukkumarkkinoi-den rajoitteet niiden rahoittaessa taseitaan voimistavat rahapolitiikan vaikutuksia
taloudessa. Tss yhteydess puhutaan tavanomaisesti rahapolitiikan vaikutusten
pankkikanavasta osana rahapolitiikan yleist vlittymismekanismia. Tss tyss
osoitetaan, ett niden rahoitusrajoitteiden vaikutukset rahapolitiikan vlittymi-
seen ovat itse asiassa teoreettisesti epselvi, vaikka tavanomaiset kotitalouksien
ja yritysten varallisuuden mrytymist selittvt teoreettiset mallit puoltavatkin
ennakkoksityst pankkikanavasta rahapolitiikan vaikutuksia voimistavana tekij-
n. Tyn empiirisess osassa tutkitaan G8-maiden tilastohavaintoja, joiden ei
katsota tukevan oletusta, ett pankkisektorin talletukset reagoisivat rahapolitiikan
kiristmiseen negatiivisesti ja voimakkaammin kuin lainananto. Tm oletus on
sen ksityksen mukainen, ett pankkikanava voimistaa rahapolitiikan vaikutuksia.
Empiirinen nytt ei kaiken kaikkiaan ole yksiselitteist, mutta viittaa siihen, ett
talletukset vaihtelevat mytsyklisesti ja jonkin verran lainanantoa vaimeammin.
Toisin kuin pankkien lainananto, pankkitalletusten kokonaismr ei myskn
nyt merkittvsti reagoivan koron muutoksiin. Niden tulosten perusteella ei ole
ilmeist, ett pankkikanava voimistaa rahapolitiikan vaikutuksia. Tm tutkimus
siis osaltaan korjaa alan kirjallisuudessa jo muutaman vuosikymmenen vallinnutta
vrinksityst pankkien merkityksest rahapolitiikan vlittymisess.
Avainsanat: luottokanava, rahapolitiikan vlittyminen, pankkien rahoitusrajoitteet
JEL-luokittelu: E44, E52, G32
-
7/29/2019 Bank Cycles 2009
7/62
5
Contents
Abstract .................................................................................................................... 3
Tiivistelm (abstract in Finnish) .............................................................................. 4
1 Introduction ...................................................................................................... 7
2 The conventional view of the bank lending channel ................................... 10
3 The role of bank reserves in monetary policy ............................................. 14
4 An alternative view: bank balance sheets as an attenuator
of monetary policy transmission ................................................................... 15
5 Evidence provided by aggregate data .......................................................... 22
6 Conclusions ..................................................................................................... 35
References .............................................................................................................. 39
Appendix 1 ............................................................................................................. 43Appendix 2 ............................................................................................................. 47Appendix 3 ............................................................................................................. 54
-
7/29/2019 Bank Cycles 2009
8/62
6
-
7/29/2019 Bank Cycles 2009
9/62
7
1 Introduction
In any developed modern economy, bank balance sheets comprise a major part of
the money stock. For example, at the end of 2006 bank deposits amounted to
83.8% of broad money (M3) in the United States, 101.3% of broad money (M3)
in the Euro Area, and 121.6% of broad money (M4) in the United Kingdom.1
Bank lending to the private sector constituted 70.6% of US broad money, 136.4%
of Euro area broad money and 150.7% of UK broad money.2 The behaviour of
bank lending and bank deposits is therefore clearly important to policy makers.
Nevertheless, despite a large volume of research, there is still no clearly
established consensus about the response of banks to changes in monetary policy.
Much of the literature adopts the following perspective. The monetary
authorities tighten monetary policy by reducing available reserves at the centralbank. This in turn induces a contraction in broader measures of money by
reducing reservable deposits ie those deposits against which reserve
requirements are calculated. This monetary contraction then operates through two
different channels.
The first is the standard interest rate channel of monetary policy. Banks are
prepared to pay more for overnight borrowing of reserves. Rates of return on non-
bank assets must also rise, in order to persuade non-banks to hold less wealth in
the form of reservable deposits. Over time these higher interest rates lower the
demand for bank loans and also reduce the cash flows of firms and companiesborrowing at short term rates of interest. As a result the rate of growth of
aggregate demand and output and the rate of inflation decline.
Many argue that a second channel, the bank lending channel, operates when
at least some banks are liquidity constrained. Following a reduction in reserves
and bank deposits, constrained banks are unable to substitute wholesale market
funding for lost reservable deposits and so must reduce their lending by more than
unconstrained banks. If some of their borrowers are also bank dependent, ie
cannot themselves substitute other non-bank sources of finance for bank loans,
then the consequence of this reduced loan supply is a larger decline in bank
lending than would result from the interest rate channel alone.
This argument for an additional role for a bank lending channel of monetary
policy was originally developed by Bernanke and Blinder (1988), amending the
standard ISLM model to incorporate an additional balance sheet induced
contraction of the supply of credit, and by Stein (1998), who develops the model
in a more rigorous fashion incorporating an adverse selection problem in the bank
1 All data in this paragraph are computed from International Financial Statistics, October 2007.2 The bank lending figures for the UK and Euro area substantially exceed the bank deposit base.
The US is not so much different once allowance is made for loans repackaged into asset backedsecurities.
-
7/29/2019 Bank Cycles 2009
10/62
8
funding market. Investigations of the bank lending channel are reported by
Bernanke and Blinder (1992), Kashyap and Stein (1995), Kashyap and Stein
(2000) and several other researchers. Kashyap and Stein (2000) conclude, from
their examination of a large panel of US bank accounting data, that amongst thesmaller banks in their sample those that have lower liquidity ratios and therefore
more likely to be constrained in their access to wholesale funding, do indeed
respond more to changes in monetary policy than banks with higher liquidity
ratios. However the results of several other empirical studies (for example
Ehrmann et al, 2003) provide only mixed support for the prediction that
constrained banks respond more to monetary policy than unconstrained banks.
Our paper reconsiders this bank lending mechanism. We show in the
analysis developed in Section 4 and Appendix 2 below that this standard
exposition of the bank lending channel is flawed, assuming as it does that
monetary tightening results in an net increase in the demand of banks for
wholesale funding. The correct logic, allowing for the possibility that bank
demand for wholesale funding may either rise or fall following an increase of
interest rate, is as follows:
(i) If higher interest rates reduce the supply of bank deposits more than the
demand for bank lending then constrained banks become more financially
constrained, their effective marginal cost of funding rises, and their loan rates
and volume of lending both decline by more than those of unconstrained
banks. The bank lending channel then amplifies the interest rate impact ofmonetary policy.
(ii) If higher interest rates reduce the demand for bank lending more than they
reduce the supply of deposits then the opposite outcome emerges: constrained
banks become less financially constrained, their effective marginal cost of
funding falls, and their loan rates and volume of lending decline by less than
those of unconstrained banks. The bank lending channel then attenuates the
interest rate impact of monetary policy.
We will further argue that this correction to the understanding of the bank lendingchannel can have significant policy implications in situations, such as the present
time, when bank funding is under strain.
We develop this argument as follows. The two sections following this
introduction provide some supporting preliminary discussion. Section 2 reviews
the principal contributions to the existing literature on the bank lending channel
(Appendix 1 provides a fuller review of the large empirical literature). We argue
in this section that the empirical results of Kashyap and Stein (2000) can be
interpreted as reflecting either an amplifying or attenuating impact of the bank
lending channel. Their work, while using a rich microeconomic data set, does not
resolve the ambiguity of the sign of the bank lending channel.
-
7/29/2019 Bank Cycles 2009
11/62
9
Section 3 discusses a departure that we make from previous theoretical
models of the bank lending channel. This is our assumption that the monetary
policy instrument is the short term interest rate, not the stock of reserves with the
central bank. We note that while reserves with the central bank (whethermandatory or not) play an essential role in the implementation of monetary policy,
it is usually more convenient, as well as more in line with actual central bank
practice, to treat reserves as we do as passively responding to the stock of
reservable deposits.
Following these preliminaries Section 4 then discusses the extension of
standard models of banking sector equilibrium to incorporate the impact of the
level of market interest rates on bank funding, ie on the flows of bank lending,
deposits, and wholesale borrowing, and hence on the margins between market and
bank interest rates. We show that the margins between bank lending rates and
market rates vary with the level of market rates when banks are financially
constrained ie when banks cannot freely access markets for wholesale borrowing;
but that the impact of market interest rates on these margins is ambiguous,
depending upon whether an increase of market interest rate produces a net
increase or decrease in bank reliance on wholesale funding. Appendix 2 provides
a full statement of our model and an extension to the case of imperfect
substitution between loan and deposits and financial constraints that affect some
banks but not others.
Sections 5 attempts to resolve the ambiguity in the sign of the bank lending
channel through an empirical examination of aggregate quarterly data for the G8OECD economies from 1970Q1 onwards (the sources of these data are described
in Appendix 3). We first report the time series behaviour of bank deposits and
loans, showing that in all countries the unconditional volatility of loan growth is
higher than that of deposits; that is to say, in periods of monetary expansion loans
grow by more than deposits while in periods of monetary contraction they fall by
more than deposits. We also conduct vector auto regression analyses of the
relationships between interest rates, bank deposits, bank lending, and nominal
GDP and report the resulting impulse response functions for the relationship
between interest rates, bank deposits, and bank lending. The results are far fromclear cut, but they are consistent with the view that lending responds more
strongly and often earlier than bank deposits to a shift in monetary policy. This
implies that, if the deposits and loans of constrained banks behave similarly to
those of the banking sector as a whole, then the bank lending channel will
attenuate the impact of monetary policy.
Section 6 summarises our findings and provides some further discussion of
the interaction between the bank lending channel and the closely related bank
capital channel. The conventional exposition of the bank lending channel
assumes that liquidity constraints do not alter in response to a shift in monetary
policy. The bank capital channel introduces the possibility that liquidity
-
7/29/2019 Bank Cycles 2009
12/62
10
constraints may become more binding following a tightening of monetary policy
and become less binding following a loosening of monetary policy. Our analysis
of the interaction between the bank lending channel and the bank capital channel
also allows us to discuss the conduct of monetary policy in periods when thebanking sector becomes critically undercapitalised, for example in Japan in the
1990s, and worldwide following the credit crunch of the summer of 2007. We
find that undercapitalisation and expectations of continued low growth of bank
income can reinforce the attenuation of monetary policy created by the bank
lending channel, leading in the most extreme cases to a deflation which monetary
policy alone is powerless to escape.
2 The conventional view of the bank lendingchannel
This section restates the theory of the bank lending channel and assesses the most
widely cited empirical study of its size, sign, and importance. The original
theoretical presentation is Bernanke and Blinder (1988). They assume that banks
hold three assets reserves, loans, and short term bonds and issue one liability
bank deposits. Loans and bonds are imperfect substitutes, both as sources of
finance to borrowers and as assets held in bank portfolios. In consequence the
stock of bank credit depends on the spread between bank loan and bond marketrates of interest.
Bernanke and Blinder then discuss the implications for monetary transmission
in an ISLM setting, arguing that a tightening of monetary policy results not only
in the standard leftward shift in the LM curve but also at the same time as bank
loan rates increase in response to the monetary policy tightening and thus reduce
the supply of investible funds to the market in a leftward shift in the IS curve.
They argue that the impact of bank balance sheets is thus to amplify the
transmission of monetary policy. They note that the IS curve will be affected by
disturbances to the supply or demand for bank credit (both of which will affectbank loan rates independently of market rates of interest) and argue that credit
stock targeting can be preferable to monetary targeting when money demand is
relatively unstable compared to credit demand.
In Stein (1998), in contrast to Bernanke and Blinder (1988), it is bank liability
management rather than bank asset management that plays a key role in monetary
transmission. This is a more formal model of the capital market frictions limiting
bank access to wholesale market funding. His banks hold two assets reserves
and loans and issue two liabilities insured deposits and wholesale market
liabilities (eg certificates of deposit). He considers a separating equilibrium
generated by adverse selection between smaller more opaque banks, unable to
-
7/29/2019 Bank Cycles 2009
13/62
11
access wholesale liabilities and therefore relying exclusively on deposits for
funding their lending, and larger transparent banks, able to access additional
wholesale liabilities freely at market rates of interest.
Stein (1998), like Bernanke and Blinder (1988), concludes that the impact ofbank balance sheets on loan supply amplifies the impact of monetary policy, but
his model predicts that this amplification will be limited to constrained banks that
are unable to substitute wholesale finance for a monetary policy induced reduction
in bank deposits. He also points out that for these constrained banks a disturbance
to bank deposits, eg an inflow of deposits financed by sale of other assets, will
affect the supply of bank credit.
A substantial empirical literature reviewed in Appendix 1 investigates the
magnitude of this bank lending channel. This literature has reached a consensus
on one point that it is extremely difficult to separately identify the impacts of
bank loan supply and bank loan demand using aggregate data. Therefore the more
informative empirical studies use individual bank accounting data. It is however
widely accepted that even with individual bank data it is a difficult to disentangle
the impact of loan demand and loan supply. As a result there is not yet a
consensus on the importance of the bank lending channel in monetary
transmission.
The most widely cited of the micro-econometric studies of the bank lending
channel is Kashyap and Stein (2000).3 Their data comprise the US call report data
available for all US commercial banks. This source provides Kayshap and Stein
(op. cit.) with nearly one million bank-observations covering 1976Q2 to 1993Q2.Their specification seeks to control for the impact of bank loan demand by
examining, in particular, differences in behaviour between large and small banks.
They apply a two-stage estimation procedure. In the first stage they estimate
cross-sectional regressions of the change in the log of bank lending, using four
lags of the dependent variable together with Federal Reserved district dummy
variable and the lagged value of the ratio of securities plus federal funds sold to
total assets (their bank liquidity ratio). In the second stage they conduct univariate
times series regressions, regressing the coefficient on the bank liquidity ratio
estimated from the first-stage cross-sections for each time period on the currentand four lagged values of a measure of the stance of monetary policy plus a linear
time trend and also (in an extended bivariate specification) on the current and four
lagged values of the growth of GNP.
3 While we reinterpret their results, we must praise the highly professional attitude taken byKashyap and Stein (2000) towards the public dissemination of their data. They went toconsiderable lengths to correct for breaks in the call report data. The documentation of their datawork together with the complete and updated US call report data is maintained on the Federal
Reserve Bank of Chicago website. This resource offers the opportunity for further research, testingdirectly the relationship between bank funding and monetary transmission.
-
7/29/2019 Bank Cycles 2009
14/62
12
They conduct this estimation procedure for two measures of bank lending,
commercial and industrial loans and total loans, and for three alternative
indicators of the stance of monetary policy: the negative of the Federal-Funds
rate; the Boschen-Mills (1995) indicator of the stance of monetary policy based onthe reading of FOMC documents; and the Bernake-Mihov (1998) indicator of the
stance of monetary policy based on a general VaR specification (in all cases a
higher value of the monetary policy indicator represents a loosening of monetary
policy.) They also estimate separately for three different size classes of banks:
small banks falling in the 095th percentile of the size distribution; middle sized
banks falling in the 9599th percentiles of the size distribution; and large banks in
the 99100th percentile, and for both their univariate (monetary policy only) and
bivariate (monetary policy plus GNP growth) second stage specifications.
Their discussion highlights the sum of the second stage coefficients on the
current and lagged indicator of the stance of monetary policy, reported in their
Table 3 page 417. They find that for large banks the sum of these coefficients is
always positive and in most cases statistically significant at the 5% or 1% level ie
following episodes of monetary policy tightening (when the monetary policy
indicators are negative) there is a negative relationship between liquidity and
growth of bank lending (so large banks with more liquid balance sheets are
revealed to have reduced their lending by more than large banks with less liquid
balance sheets). For small banks they obtain precisely the opposite result, the sum
of the second stage coefficients on the indicator of the stance of monetary policy
is always negative and mostly statistically significant. For both large and smallbanks these coefficient estimates are larger and of even greater statistical
significance in the bivariate version of the second stage specification, including
also current and lagged GNP growth. For commercial and industrial lending the
difference in the sum of coefficients between small and large banks, is always
statistically significant at the 1% level. For total lending the difference in
coefficients is statistically significant at this level only in the bi-variate
specification, not in the univariate specification.
Kashyap and Stein (2000) summarise their results as follows (page 425):
Within the class of small banks changes in monetary policy matter much morefor the lending of those banks with the least liquid balance sheets. They place
particular emphasis on the major coefficient differences between large and small
banks, on the grounds that loan supply effects should be relatively unimportant for
large banks and hence that the difference in estimates between small and large
banks are a relatively clean estimate of loan supply impacts uncontaminated by
the impact of loan demand. They conclude that their analysis reveals strong
empirical support for a statistically and quantitatively important link between
bank liquidity and bank loan supply of the kind set out in Stein (1998) ie that the
bank lending channel does amplify monetary policy transmission.
-
7/29/2019 Bank Cycles 2009
15/62
13
Do the findings of Kashyap and Stein (2000) resolve the question of whether
the bank lending channel amplifies or attenuates the impact of monetary policy?
The answer is no, they do not. Their findings, while consistent with the view that
small financially constrained banks are forced to respond more to changes inmonetary policy than are small financially unconstrained banks, can readily be re-
interpreted as reflecting an entirely different mechanism. Many small banks, for
reasons of either commercial interest or concern for the welfare of their
customers, offer implicit hedging of loan interest rate risk. Such banks will
maintain narrow interest margins and offer their customers relatively cheap
lending facilities during periods of tight monetary policy, in return for offering
somewhat less generous lending rates, and operating with relatively high interest
margins, during periods of loose monetary policy. Only small banks with strong
community presence and ongoing rather than transient customer relationships can
operate in this way. Such banks can be expected also to hold a relatively high
proportion of liquid assets, to permit the resulting variability of their net interest
income over time. This is a simple alternative explanation of the Kashyap and
Stein (2000) finding that more liquid small banks respond less to monetary policy
that less liquid small banks, but this explanation has nothing to do with the impact
of financial constraints of the bank lending channel. Their research, while
providing valuable insight into bank behavior, provides no conclusive evidence on
the sign of the bank lending channel.4
Can the results of Kashyap and Stein (2000) and other related studies be
interpreted as arising because liquidity constraints bind more tightly duringperiods of monetary contraction and less tightly during periods of monetary
expansion? The empirical tests conducted in this literature are inappropriate for
testing this hypothesis. If monetary policy is amplified by a tightening of bank
liquidity constraints then the largest impact should not be observed for the
smallest, most illiquid, and least capitalized banks, since these are already the
most severely liquidity constrained, but for banks at intermediate levels of size,
liquidity, or capitalization for whom liquidity constraints begin to bind.
4 Kashyap and Stein (2000) were aware of this possible bias. On pages 415416 they note that thedifference in coefficient estimates for small banks could reflect not amplification of monetary
policy via the bank lending channel but heterogenous risk aversion, with some relativelyconservative banks responding less to monetary policy and at the same time operating a moreliquid balance sheet, in comparison to other less conservative banks. Their defense against thiscriticism is to argue (a) that there is another bias running in the opposite direction, with some
banks with relatively cyclically sensitive portfolios also preferring to hold a more liquid balancesheet (b) with the further maintained assumption that large banks can never be liquidityconstrained, their finding that the lending of large liquid banks respond more to monetary policythan that of large illiquid banks shows that the second of these two biases dominates the first.However this does not settle the matter since it is possible that even large banks can be liquidity
constrained. In this case the observed coefficients estimates for large banks could arise becausethere is an attenuation of the monetary policy response of less liquid large banks.
-
7/29/2019 Bank Cycles 2009
16/62
14
3 The role of bank reserves in monetary policy
This section explains our decision to assume that the instrument of monetary
policy is the short term interest rate, with required reserves responding passively
to changes in deposits. Reserve requirements do not matter for the transmission of
monetary policy, at least in developed financial systems. They are effectively a
tax on commercial banking operations, with the tax rate depending on the extent
that they are unremunerated, the level of short term interest rates, and the excess
reserves they lead banks to hold over what they would hold from commercial
prudence. But the level of this tax is too small to make much difference to banks
choice of assets and liabilities and in any case central banks in developed financial
markets do not alter required reserve ratios as part of their conduct of monetary
policy. Banks determine their lending and funding decisions according to theirview and those of the financial markets on the future course of market interest
rates; and also on the ability of banks to obtain finance from these markets and the
terms on which this finance is made available.
In our analysis of monetary transmission we could assume that either the
supply of reserves or the short term rate of interest is the policy instrument. The
level of short term policy rates r determines the demand for bank deposits D(r)
and hence, for a required reserve ratio of, the demand for reserves R(r) = D(r)
with the central bank. Provided this function R(r) is one-to-one mapping then any
monetary policy can be described either in terms of movements in reserves or ofshort term interest rates. It a matter only of modelling convenience which
approach is used. Applied correctly both must yield the same answer.
The choice of monetary instrument, between a monetary stock (whether the
monetary base or some broader measure of money) or an interest rate (whether
overnight or at some longer time horizon), is not just a matter of convenience in a
stochastic setting. If economic relationships are disturbed by shocks that arrive
after the decision over the monetary policy instrument is made, then one
instrument may be more successful than another in achieving the ultimate policy
target (see Poole, 1970). This important insight does not matter to our analysis,
since our model is, for the sake of tractability, deterministic.5
5 The supply of reserves by the central bank also plays a further important role in stressed marketsituations, for example following those following the 9/11 terrorist attacks or in the wake of thecredit crisis of the summer of 2007, which produce a sudden and unanticipated increase in demandfor safe liquid assets. In these situations the central bank must by a classic lender of last resortoperation provide additional liquidity in order to maintain confidence in bank liabilities. Suchliquidity shocks do not need to be modelled in order to understand the role of banks in monetary
transmission, for the central bank response is explicitly intended to prevent any change inmonetary policy stance.
-
7/29/2019 Bank Cycles 2009
17/62
15
4 An alternative view: bank balance sheets as anattenuator of monetary policy transmission
This section examines how changes in monetary policy, operating through the
policy interest rate (denoted by r), affect both the volumes of bank loans, deposits
and wholesale funding (denoted by L, D and W) and bank interest rates on these
assets and liabilities (denoted by rL, rD, and rW). Banks compete with financial
markets on both sides of their balance sheets and so a change in the market rate of
interest directly affects both the demand for bank loans and the supply of deposits
and wholesale funds. We show that in this case bank balance sheet constraints can
attenuate monetary policy transmission.
Our aim is to extend an otherwise standard model in order to analyse the
impact of the aggregate level of interest rates on bank funding (Appendix 2
provides a fuller exposition). For this task it is appropriate to work with a reduced
form model in which the interest rate coefficients reflect both the direct and
indirect impacts of monetary policy on bank funding flows (the direct impacts are
the standard interest elasticities for the demand for lending and the supply of
deposit and wholesale funding; the indirect effects are those additional impacts of
interest rates on bank funding that arise for a number of other reasons, for
example improved cash flows or net worth of bank borrowers or changes in
expectations about future aggregate demand and output, leading customers to
borrow more from banks or to increase their bank deposits). Empirical evidencesuggests that the direct response of bank loan demand to long term interest rates is
rather small, but this is consistent with our analysis.
Since we are concerned only with the impact of small movements of interest
rates we can assume constant interest-semi elasticities (exponential functions) of
aggregate reduced form demand for lending and aggregate reduced form supply of
wholesale and insured deposit funding, allowing us to use a version of the
standard Klein-Monti model of banking competition to model the impact of the
policy rate r on the banking sector.6
There are N identical banks, indexed by n, in Cournot quantity competition.As in Stein (1998), we distinguish four assets and liabilities in the balance sheet
identity for bank n7
nnnn WDRL +=+ (4.1)
6 Klein (1971), Monti (1972). We follow closely the exposition in Freixas and Rochet (1997)chapter 3.7 This focus on liability management and the use of wholesale funding is not material to our
argument, we could equally well adopt the Bernanke and Blinder assumption that the banks holdliquid assets (bonds) and make no use of wholesale borrowing.
-
7/29/2019 Bank Cycles 2009
18/62
16
These assets and liabilities are
Ln = loans
R
n
= non-interest bearing required reserves, responding passively to the reserverequirement according to Rn = Dn
Dn = insured deposits
Wn = other funding, including equity capital and wholesale debt finance.
We assume the following equations for the aggregate volumes of banking sector
loans, deposits, and wholesale finance
))rr(exp()rexp(L)r,r(LL LLL0
L
N
1n
n == = (4.2a)
))rr(exp()rexp(D)r,r(DD DDD0
D
N
1n
n == = (4.2b)
))rr(exp()rexp(W)r,r(WW WWW0
W
N
1n
n == = (4.2c)
The loans of different banks are perfect substitutes for each other ie borrowers
may be bank dependent but they are not dependent on individual banks. Similarly
there is perfect substitutability between the deposits of different banks and also
between the wholesale liabilities of different banks. Perfect substitutability
implies, as shown in equations (4.2a, 4.2b, 4.2c), that aggregate assets and
liabilities are functions of the monetary policy rate r and sector wide bank interest
rates on loans, deposits and wholesale funding.
Banks then seek to maximise one period profits (net interest margins)
n
W
n
D
n
L
n WrDrLr = (4.3)
taking account of the impact of their borrowing and lending on bank interest rates
(through (4.2a), (4.2b), (4.2c)) and subject to the balance sheet constraint (4.1)
and the reserve requirement Rn = Dn.
Before discussing the equilibrium outcome, we comment on the key
parameters. Three elasticity parameters L, D and W capture thesubstitutability of bank and non-bank assets and liabilities. The parameter Lreflects the extent to which borrowers are bank dependent. In the limiting case
L borrowers can perfectly substitute other sources of finance eg commercialpaper or bond issuance, and bank loan rates have no impact on borrower credit
and money transmission. As we establish shortly, a necessary condition for the
existence of a bank lending channel is that the sum of these elasticities is finite:
L + D + W < , ie not only are bank borrowers bank dependent but also banks
-
7/29/2019 Bank Cycles 2009
19/62
17
themselves face rising costs of funding as they expand deposit and wholesale
liabilities.
These parameters determine how the volume of aggregate banking sector
assets and liabilities respond to the various margins between the market interestrate r and the rates on bank instruments. The standard Klein-Monti model is a
limiting case where W+. In this case wholesale funding to the banking sectoris available in infinitely elastic supply (no banks are financially constrained) and
can fully substitute for any variations in the supply of deposits (ie change in D0)
or fully match any variation in the demand for bank lending (ie a change in L0).
We further amend the standard Klein-Monti model by introducing a monetary
policy (market interest rate) impact on aggregate bank assets and liabilities,
through the parameters L, D and W. We think of these as a shorthand for theimpact of market interest rates (monetary policy) on aggregate, economy wide,
flow of funds, both for retail investors and wholesale market participants. A
structural interpretation of our model suggests a priori that these three parameters
all have positive signs. In the event of a monetary policy tightening the
intertemporal shift of consumption and investment expenditures (with reduced
expenditure today and higher expenditure in future periods) can be expected to
lead to a decline of both bank and market financing and an increase in both retail
and wholesale portfolios. However it is possible that in reduced form the indirect
impact of a tightening monetary policy could lead to an net decline of bank
deposits ie the sign ofD and W is ultimately an empirical matter. The relative
magnitude of these latter parameters is not critical, although we expect in a smallopen economy that W
-
7/29/2019 Bank Cycles 2009
20/62
18
These two margin equations, together with the banking sector balance sheet
identity (4.1), determine the three endogenous bank interest rates and hence
banking sector equilibrium.
The limiting case in which W+, provides the benchmark for assessing boththe sign and magnitude of the bank lending channel. In this special case banksaccess an infinitely elastic supply of wholesale funding for bank lending and
hence bank wholesale rates always equal short term market rates (rW = r) and both
bank loan rates and bank deposit rates move in line with market rates. Since the
margins between bank loan and deposit rates and market rates are unaffected by
monetary policy, bank balance sheets then neither amplify nor attenuate the
impact of monetary policy, and the reduction in bank lending is determined solely
by the demand for bank lending as determined by the aggregate interest rate loan
demand elasticityL.We compare against this benchmark the response of bank lending to changes
in monetary policy when banks have only limited access to wholesale finance and
hence are liquidity constrained. When banks are liquidity constrained then there is
a positive shadow price of internal funds)(N
1
)(N
1r
WWLL
W ++
+= .
Assuming an exponential supply of wholesale funding this shadow price is given
(see Appendix 2) by
)N)W/W(ln()( 10n1Ww ++= (4.5)
ie this shadow price depends on the interest rate elasticity of the supply of
wholesale funding, on the volume of wholesale funding utilized relative to what is
available, as well as on the number of banks competing for wholesale funds. In
line with the previous literature on the bank lending channel we assume that
liquidity constraints are not directly affected by shifts of monetary policy ie that
W0, W, and W all remain constant. Section 6 discusses the possibility thatmonetary policy may directly affect liquidity constraints for example a tightening
of monetary policy might alter wholesale funding available to banks W0, perhaps
because of an impact of monetary policy on bank capital.Equation (4.2a) indicates that, relative to the standard benchmark, the
direction of the bank lending channel depends on the direction of change of the
interest rate margin rLr .If this margin increases, following a monetary policy
tightening, then the bank lending channel amplifies the impact of monetary policy
on bank lending (the Bernanke-Blinder, 1988; and Stein, 1998, prediction.) If this
margin decreases then the impact on bank lending is attenuated.
The response of this margin depends upon a simple condition that emerges
directly from the solution of the model. Consider an increase in short term market
rates of interest ofr. From equations (4.4a) and (4.4b) it is apparent that all bank
-
7/29/2019 Bank Cycles 2009
21/62
19
interest rates move togetherrL = rD = rW: As shown in Appendix 2, in order to
maintain the aggregate banking sector balance sheet constraint, the response of
these banking sector interest rates to a change in monetary policy then must
satisfy
1W)1(DL
W)1(DL1
r
r
r
r
r
r
WDL
WDLWDL 0. In this case
loans decline by LL while deposits (net of reserves) increase by DD(1) andwholesale bank funding increases by WW. But this leaves banks with an excessof funding (loans have declined while both deposits and wholesale funding have
increased) so this cannot be an equilibrium (the only exception is if one of these
three assets and liabilities is in infinitely elastic supply in which case a small
change in an interest margin can restore the banking sector balance sheet
constraint.) In order to restore the banking sector balance sheet constraint, thebanks must absorb this additional funding, and to do this they lower the various
bank interest rates in order to generate greater loan demand and reduce the
volume of deposit and wholesale finance. This excess funding available to banks
when monetary policy is tightened leads to a reduction in the bank loan market
rate of interest margin rLr and this in turn explains why the bank lending channel
can be expected to attenuate, rather than amplify, the impact of monetary policy
on bank lending.
The Bernanke and Blinder (1988) and Stein (1998) prediction that the bank
lending channel amplifies the impact of monetary policy on bank lending emergesunder alternative parameter assumptions, when8
0W)1(DL WDL
-
7/29/2019 Bank Cycles 2009
22/62
20
If our analysis is correct, and the sign of the bank lending channel is indeed
theoretically ambiguous, then there must be expositional mistakes in these
previous models. With the additional insight provided by our own model, these
mistakes are easily located
Bernanke and Blinder (1988) argue that since bank loan rates rise when
interest rates are tightened the IS curve must be shifted to the left by a
tightening of monetary policy. This misleads because it does not recognize
that in the standard ISLM model, when bank balance sheets have no impact
on monetary policy, bank loan interest rates move in line with market rates of
interest. If bank loan rates respond more than one for one to changes in
market rates of interest then the IS curve moves to the left following a
monetary tightening. If, as we argue is more likely to be the case, and is
undeniably possible, bank loan rates respond less than one for one to market
rates of interest then the IS curve moves to the right and the impact of
monetary policy is attenuated.
Stein (1998) argues that following a monetary tightening constrained banks
are unable to replace a loss of deposits with wholesale finance and hence must
reduce their lending by more than unconstrained banks. This argument is
flawed, relying as it does on the unstated assumption that banks lose more
deposits than loans following a monetary tightening. .If as we argue there is a
deposit inflow following a tightening of monetary policy there is instead a
reduction of demand for wholesale funding by constrained banks and onceagain the impact of monetary policy is attenuated.
We have reached our conclusions using a very simple model of banking
equilibrium. Does the same conclusion still hold under alternative specifications
of banking competition? A problem with using the basic Klein-Monti
specification is that all banks are funding constrained to the same degree.
Appendix 2 explores a more general model, relaxing the Klein-Monti assumption
that the deposits (and the loans) of different banks are perfect substitutes, and
instead assuming imperfect substitution between the loan and deposit products ofdifferent banks, with elasticities of substitution ofL and D.
This second specification allows us to develop a version of the model
corresponding more closely to Stein (1998), with two groups of banks one
constrained the other unconstrained. With the assumption that L > 0 we reach thesame conclusion as before ie that the impact of financially constraints on banks is
to attenuate the impact of monetary policy on bank lending. The Stein (1998)
results are restored only if, following a rise of interest rates, there is a deposit
outflow from financially constrained banks that exceeds the decline in demand for
their lending.
-
7/29/2019 Bank Cycles 2009
23/62
21
In our model the margins between bank interest rates (the right hand side of
the first order conditions (4.4a) and (4.4b)) remain constant. There are several
reasons for thinking that bank loan supply might increase and loan margins
contract in a monetary policy expansion (or bank loan supply might reduce andloan margins increase in a monetary policy contraction.) Reasons for expecting
such a loan supply amplification of monetary policy include the following
(i) a monetary policy expansion might promote greater competition between
banks, and a narrowing of bank margins, as they attempt to claim an
increased share of an expanding market.
(ii) structural shifts in banking competition can also have monetary policy
implications. For example, the liberalization of banking markets and the
rapid innovation in structured credit markets since the mid-1980s has
increased bank access to wholesale funding, thus reducing interest margins
and contributed to relatively rapid growth of the stock of bank credit. It is
possible that monetary policy might affect the pace of such structural
change, by reducing the costs of bank funding and increasing loan supply
in a monetary expansion.
(iii) monetary policy might affect banks assessments of their portfolio risks.
Following a monetary policy expansion banks may consider default less
likely and be willing to lend at lower interest margins.
Such loan supply effects may play an important role in monetary transmission.Such supply shifts also help to explain why it is relatively difficult to predict the
response of bank lending to changes of interest rates. But these loan supply effects
do not alter our main finding, that financially constrained banks will respond less
than unconstrained banks to shifts in monetary policy if a monetary tightening
leads to a net inflow of funding (lending falling more than deposits) and so
reduces their funding requirements.
There is however a possibility that a monetary policy tightening might reduce
the book net worth or the market capitalization of financially constrained banks
and hence reduce the amount of wholesale funding available (at a given cost offunding rW). This bank capital channel could in some circumstances
substantially reduce W0 in equation (4.5) have a major impact on the supply of
bank loans. This mechanism might offset or even reverse the impact of the bank
lending channel as modeled in this section. We discuss this possibility in our
concluding Section 6.
-
7/29/2019 Bank Cycles 2009
24/62
22
5 Evidence provided by aggregate data
This section examines the response of bank deposits and lending to monetary
policy changes, as revealed by aggregate data on bank lending, bank deposits,
GDP and interest rates for the G8 countries.9
Figure 1, panels (a)(h), present aggregate macroeconomic and banking data,
detrended using the Hoderick-Prescott filter. There is less consistency of
definition for these banking variables across countries than for the
macroeconomic variables. The banking data are also affected by both reporting
problems and by structural changes in the banking industry. Nonetheless some
conclusions can be drawn, about whether monetary policy increases or reduces the
demand by the banking sector for wholesale funding, and hence on the likely sign
of the bank lending channel
For the seven countries where a comparison can be made (we do not have a
long enough time series for the UK) sight deposits fluctuate very much more
than total bank deposits. This suggests that, while it may be true that a
tightening of monetary policy can reduce holdings of sight deposits, most of
this impact is a shift between sight and time deposits with a relatively small
overall effect on bank wholesale funding requirements.
For all eight countries bank lending fluctuates pro-cyclically and with these
movements either co-incident, or (eg in the United States) slightly leadingmovements in real and nominal GDP.
In six of the eight countries (the exceptions are France and Italy) the cyclical
movements in total bank deposits occur at about the same time and in the
same direction as the movements in nominal and real GDP, with some
indication eg for the US that the deposit movements lag those of bank
lending. In the United States, the United Kingdom, Australia and Canada (but
not in Germany or Japan) the amplitude of the cyclical deposit movements
appear to be somewhat less than those of total bank deposits. This suggests
that movements in bank funding requirements over the business cycle,
whether induced by monetary policy or by aggregate shocks, are procyclical
and relatively small. If the bank funding movements of liquidity constrained
banks are similar then the bank lending channel will have a relatively small
attenuating impact on monetary transmission over the course of the business
cycle.
9
Our data sources, together with a number of adjustments to the banking data to correct for bothreporting breaks and seasonality, are described in Appendix 3
-
7/29/2019 Bank Cycles 2009
25/62
23
Figure 1. Aggregate data, HP trend adjusted
Panel (a) United States
Panel (b) Germany
-
7/29/2019 Bank Cycles 2009
26/62
24
Panel (c) United Kingdom
Panel (d) France
-
7/29/2019 Bank Cycles 2009
27/62
25
Panel (e) Italy
Panel (f) Japan
-
7/29/2019 Bank Cycles 2009
28/62
26
Panel (g) Australia
Panel (h) Canada
-
7/29/2019 Bank Cycles 2009
29/62
27
In both France (during the 1980s and early 1990s) and in Italy (during the
1990s), there are substantial differences in the cyclical movements in bank
deposits and in bank lending, indicating that in these two countries during
these periods there have been some relatively large shifts in bank fundingrequirements. But there is no obvious consistent pattern to these movements
and in the case of France the overall magnitude of the cyclical fluctuations
again appears somewhat smaller for deposits than for loans.
These charts also display the movements in a measure of real interest rates
(the nominal market rate of interest less the rate of growth in the smoothed
nominal GDP from the Hoderick-Prescott Filter). This also moves pro-
cyclically, indicating that it is difficult to disentangle the impact on bank loans
and deposits of interest rates from the impact of cyclical movements in
nominal and real output. While deposits typically do rise during periods when
real interest rates are rising, we cannot say simply from visual inspection
whether this is due to an interest rate impact or simply an increase in the
demand for deposits driven by real and nominal income growth. Our prior,
that monetary tightening reduces the level of deposits, cannot be tested from
this visual inspection.
As an alternative way of comparing the magnitude of cyclical fluctuations, we
estimated, allowing for correlation between the error terms, the following
bivariate regressions for aggregate bank loans and total bank deposits
L33
2211332211
Dln
DlnDlnLlnLlnLlnLln
++
++++=
(5.1a)
D33
2211332211
Lln
LlnLlnDlnDlnDlnDln
++
++++=
(5.1b)
Table 1 reports the resulting standard errors (for changes in the log) of both
aggregate bank deposits and aggregate bank loans obtained from these regressions
for the eight countries. Single period standard errors (rows (1) and (2)) are similarfor loans and deposits, but in order to examine cyclical volatility it is more
appropriate to look at the unconditional standard errors, allowing for the impact of
the estimated dynamics from the lagged dependent variables and correlation of
residuals. Comparison of these unconditional standard errors (whether single
equation, reported in rows (3) and (4) or joint equation, reported in rows (5) and
(6)) provides an alternative quantitative comparison of the magnitude of cyclical
fluctuations in deposits and lending. This comparison is consistent with the visual
evidence provided by Figure 1, with a much greater level of cyclical volatility in
-
7/29/2019 Bank Cycles 2009
30/62
Table1.
Stan
dar
der
rorsan
dgrangercausa
lityte
sts
from
bivar
iateregress
ions
Standard
errors(%
)
US
GER
UK
FR
IT
JA
AU
CA
Oneperiod
Loans
1.0
6
0.6
6
1.7
1
1.5
1
1.51
0.8
1
1.1
1
1.4
9
Deposits
0.9
3
1.0
4
1.7
1
1.3
3
2.02
1.2
4
1.8
3
1.3
0
Unconditional
Loans
3.2
4
8.8
4
5.3
8
4.1
2
3.83
5.4
4
5.0
0
5.3
6
Deposits
1.0
7
2.5
7
2.3
9
1.3
3
3.49
3.5
7
2.8
8
1.7
1
Bivariate
unconditional
Loans
2.8
5
13.7
5
3.4
4
3.1
1
2.63
3.2
2
3.1
7
3.7
1
Deposits
1.0
1
2.4
2
2.2
9
1.2
6
3.36
3.4
3
2.8
3
1.7
6
Granger
pvalues
Depositstolending
0.0
21
0.0
26
0.0
01
0.1
28
0.19
4
0.0
09
0.0
03
0.6
64
Lendingtodeposits
0.2
99
0.0
01
0.0
00
0.9
74
0.04
0
0.0
00
0.0
00
0.0
03
Theuniva
riateunconditionalstandarderrorsareestimatesof
)
D,
D,
D/L
L(E
3
2
1
and
)
L,
L,L/D
D(E
3
2
1
.Thebivariateunconditionalstandard
errorsare
estimatesof
)L
L(E
and
)D
D(E
.Thesemeasuretheexpe
ctedvarianceofLandDaroun
dtheirunconditionalmeans.
-
7/29/2019 Bank Cycles 2009
31/62
29
the growth rate of bank lending than in the growth rate of bank deposits, for all
countries except Italy and (when comparing joint unconditional standard errors)
Japan.
Table 1 also reports p-values for a Granger causality test, for both the impactof lagged deposits on lending growth and of lagged lending on deposit growth.
These show that in six of the eight countries lagged growth of lending is
statistically significant (at the five per cent level) in predicting the growth of bank
deposits while in four of the eight countries lagged deposits are statistically
significant in predicting the growth of bank lending. However, while providing
some evidence that changes in bank deposits may affect bank lending, this
bivariate specification does not distinguish a relationship arising because of
balance sheet funding constraints from a shock affecting bank customer demand.
Finally, in order to investigate the impact of macroeconomic developments on
bank balance sheets, we estimated a vector auto regression model for all eight
countries using the sample period 1975q12007q2 with four variables: the level of
nominal interest rates (i), and the rates of growth of aggregate bank deposits
(lnD), aggregate bank lending (lnL), and of nominal GDP (lnYn), and threelags10
D
n
3 )i,Yln,Dln,Lln)(L(aDln += (5.2a)
L
n
3 )i,Yln,Dln,Lln)(L(bDlnLln ++= (5.2b)
L
n
3
n )i,Yln,Dln,Lln)(L(cLlnDlnYln +++= (5.2c)
L
n
3
n )i,Yln,Dln,Lln)(L(dYlnLlnDlni ++++= (5.2d)
and with orthogonal shocks. ie our VaR has the ordering lnD, lnL, lnYn, i.11Figure 2 shows the impulse response functions obtained from this vector
autoregression. These show a statistically significant response of bank lending to
monetary policy in five of the eight countries (the United States, Germany,
France, Australia, and Canada, but a statistically significant response of total bank
deposits to monetary policy only in Italy There is a response on the 95% threshold
in both Germany and Australia, but in the case of Germany this is quickly
reversed (the VaR estimates appear to have imaginary roots) and in Australia the
10 We avoided using the first five years of the sample because of the very substantial negative realinterest rates found (in Figure 1) for 19701974, suggesting that the relationship between bank
balance sheets, nominal GDP, and interest rates cannot have been structurally stable over this earlyperiod.11 In any VaR analysis the results can be very sensitive to the ordering of variables. We found itdifficult to get any meaningful results if the interest rate i preceded any of the other variables (for
example if i precedes lnL we found that a positive shock to interest rates resulted in animmediate rise in bank lending, a result which presumably reflects a monetary policy reaction torising bank lending not a structural relationship), but provided the interest rate i appears last in the
ordering, changing the order of the remaining variables made relatively little difference to theresults.
-
7/29/2019 Bank Cycles 2009
32/62
30
response occurs after a long lag. In the case of both Japan and the United
Kingdom there is no evidence of any statistically significant response of the
banking sector to monetary policy at all. Thus for only one of the eight countries
Italy is characterization of the behaviour of aggregate bank balance sheetsconsistent with a tightening of monetary policy increasing bank funding
constraints.
The impulse responses also report the response of bank lending to a shock to
total deposits. On a one-period basis this appears consistent with a bank lending
channel, because a shock to bank deposits results in a one period increase in bank
lending, but this is not sustained and in any case could be interpreted as a demand
shock to lending being financed out of increases in the entire balance sheet, from
deposits as well as wholesale sources of funding. Finally we report the impact of
interest rates on the growth of nominal incomes. In the four English speaking
countries, the United States, the United Kingdom, Australia, and Canada there is a
statistically significant reduction in the growth of nominal GDP following a shock
to interest rates, after a lag of two to three quarters.
These vector autoregressions do not provide a clear-cut answer as to whether
a tightening of monetary policy increases or reduces bank funding constraints, but
they are consistent with our inspection of the detrended aggregate data and the
standard errors from our estimated bi-variate regressions. Of the eight G8
countries, only for Italy do we find any evidence that aggregate bank deposits
respond more than aggregate bank lending, either over the business cycle or in
response to a shift of monetary policy. In most countries the behaviour ofaggregate bank deposits over the business cycle can be explained by
contemporaneous movements of nominal and real GDP. While this means that a
tightening of monetary policy can reduce banking sector deposits, bank lending
falls at the same time and the overall effect of monetary policy does not appear to
be large enough to create a substantial net increase in the demand for bank
wholesale funding.
-
7/29/2019 Bank Cycles 2009
33/62
31
Figure 2. Orthogonalised impulse response functions
Panel (a) United States
Panel (b) Germany
-
7/29/2019 Bank Cycles 2009
34/62
32
Panel (c) United Kingdom
Panel (d) France
-
7/29/2019 Bank Cycles 2009
35/62
33
Panel (e) Italy
Panel (f) Japan
-
7/29/2019 Bank Cycles 2009
36/62
34
Panel (g) Australia
Panel (h) Canada
.004
.002
0
.002
0 2 4 6 8
lenddepslag, cairate, cadllend
95% CI orthogonalized irf
step
Graphs by irfname, impulse variable, and response variable
.002
0
.002
0 2 4 6 8
lenddepslag, cairate, cadldeps
95% CI orthogonalized irf
step
Graphs by irfname, impulse variable, and response variable
.005
0
.005
.01
0 2 4 6 8
lenddepslag, cadldeps, cadllend
95% CI orthogonalized irf
step
Graphs by irfname, impulse variable, and response variable
.003
.002
.001
0
.001
0 2 4 6 8
lenddepslag, cairate, cadlynom
95% CI orthogonalized irf
step
Graphs by irfname, impulse variable, and response variable
-
7/29/2019 Bank Cycles 2009
37/62
35
6 Conclusions
It has been widely accepted that constraints on the wholesale funding of bank
balance sheets amplify the transmission of monetary policy through what is called
the bank lending channel. We show that the effect of such bank balance sheet
constraints on monetary transmission is in fact theoretically ambiguous, with an
amplifying effect only when a tightening of monetary policy leads to a net
outflow of funds, ie deposit outflow exceeding the decline of lending, in
constrained banks. But if a tightening of monetary policy leads to a net inflow of
funds into constrained banks then the bank lending channel attenuates the impact
of monetary policy.
We examine macroeconomic data for the G8 countries and find no evidence
that banking sector deposits respond negatively and more than lending totightening of monetary policy, as the accepted view of the bank lending channel
requires. The overall picture is mixed but these data generally suggest that
deposits fluctuate procyclically and somewhat less over the business cycle than
bank lending, and that total bank deposits, unlike bank lending, show little direct
response to changes in interest rates. Shifts in monetary policy thus affect bank
lending more than they affect bank deposits, and therefore, to the extent that
banks are balance sheet constrained, the bank lending channel will attenuate not
amplify the response of bank lending to monetary policy. While it is possible that
the asset and liability response for some individual banks might be different, (forexample a bank whose loan business is focused in markets that are relatively
interest rate insensitive), the evidence we have examined suggests it is very
unlikely that the bank lending channel amplifies monetary policy when the entire
banking sector, and not just a section of it, is considered. Our paper has thus
corrected a misunderstanding about the role of banks in monetary policy
transmission that has persisted in the literature for some two decades.
The principal argument that can be made against this conclusion is as follows.
Our analysis shows that, in theory, the bank lending channel can either amplify or
attenuate the transmission of monetary policy. However the balance of available
empirical evidence suggests that in practice a tightening of monetary policy
reduces the amount of available to banks from wholesale markets and the
resulting increase in liquidity constraints will amplify the impact of monetary
policy on bank lending. Therefore, while the bank lending channel may not
operate exactly as described in the literature, it still provides a reasonable and
tractable characterisation of monetary policy transmission.
It may be true that a tightening of monetary policy, on at least some
occasions, increases bank liquidity constraints, but this is not the mechanism
explored in the literature on the bank lending channel. To avoid confusion this
should therefore be called something different. We believe that the appropriate
-
7/29/2019 Bank Cycles 2009
38/62
36
name is the bank financial accelerator since the underlying idea is essentially the
same as in the corporate financial accelerator, that deterioration in net worth (or
other measures of balance sheet strength) leads to a reduction in both assets and
liabilities.In terms of our theoretical model presented in Section 4, we can think of the
bank financial accelerator as a tightening of monetary policy reducing W0 (the
supply of bank wholesale funding) in our equation (4.5). We can see from
equations (4.4) and (4.5) that such a reduction in W0 will then raise the shadow
price of internal funds, increase the margin between bank interest rates and market
rates, and hence both reduce the stock of bank lending L and increase the stock of
bank deposits D.
Even when allowing for this mechanism, our analysis still has important
implications for the conduct of empirical tests for the role of bank balance sheets
in monetary transmission. Many empiricial papers have followed the example of
Kashyap and Stein (2000), seeking to test for the presence of the bank lending
channel by interacting various bank balance sheet measures (size, liquidity,
capitalisation, either measured directly or using these variables to allocate banks
into different categories) with a measure of the stance of monetary policy. These
papers have tested the hypotheses that smaller, less liquid, and less well
capitalised banks respond relatively more than other banks to changes in monetary
policy. Our analysis implies that the bank lending channel may play a role in
monetary policy transmission even when this hypothesis is rejected (as it often is,
the literature is far from reaching any consensus about the sign and magnitude ofsuch bank-balance sheet monetary policy interactions).
The appropriate way to test for the presence of the bank lending channel
arising because of differences in behaviour between constrained and
unconstrained banks is to examine the hypothesis that banks constrained by
reasons of size, liquidity, or capitalisation exhibit a relatively strong relationship
between deposit inflows and loan growth compared to other unconstrained banks.
If this is the case then there will be an amplifying or attenuating impact on
monetary policy transmission according to whether a monetary tightening reduces
or increases deposits at these constrained banks.A quite different test is needed to test for the presence of a bank financial
accelerator. Here it must be shown that a decline in measures of bank size,
liquidity, or book or market capitalisation results in an increase in liquidity
constraints and a decline of bank lending.
To conclude our paper we briefly discuss the literature on the bank capital
channel and argue that, just like the bank lending channel, the bank lending
channel yields no clear theoretical prediction about transmission of monetary
policy. In at least some circumstances it attenuates rather than amplify monetary
policy ie the prevailing view that bank balance sheets amplify the impact of
monetary policy transmission must still be rejected. We argue moreover that this
-
7/29/2019 Bank Cycles 2009
39/62
37
point is important to formulating an appropriate policy response to the current
credit crunch affecting the global banking system.
There is a fairly extensive literature on the bank capital channel. Theoretical
contributions, for example Holstrom and Tirole (1997), show that the availabilityof bank capital can limit the extent to which banks provide monitored credit. But
these analyses assume that bank capital is exogenous; they do not incorporate any
link between monetary policy and bank capital. In other contributions (for
example Diamond and Rajan (2001), bank capital is endogenous. but these
models explain long term desired capitalisation, not the effects of monetary policy
on bank capital and the resulting impact on bank loan supply.
Perhaps most relevant is the now fairly extensive literature on the dynamics of
bank capital buffers, protecting the bank against the potential costs of illiquidity
and recapitalisation.12 Banks hold a margin of capital as a buffer for absorbing
fluctuations in earnings and asset values resulting from monetary policy,
macroeconomic, market or other developments. In normal periods of operation
these fluctuations will be within the expected range of outcomes, in which case
bank capital will not then much affect the supply of bank lending. But
occasionally such losses may be much bigger than anticipated and it is then that
the availability of bank capital is likely to affect the supply of lending.
The bank capital channel can create pro-cyclical fluctuations in bank lending.
In a recession, when bank capital measured on either a balance sheet or market
value basis falls below desired levels then this can trigger a reduction in
available wholesale funding which in turn reduces the supply of bank lending. It isalso now well understood that countercyclical changes in bank capital
requirements (for example those that are introduced by the Basel II accord) may
exacerbate these cyclical fluctations in bank loan supply.13 But the impact of
monetary policy on bank loan supply via the bank capital channel, just like the
impact of bank monetary policy via the bank lending channel, is theoretically
ambiguous.
The bank capital channel impact of monetary policy is ambiguous because
bank net worth, whether measured in book or market values, may either rise or
fall following a loosening of monetary policy. This point is obvious for bookvalue capital. A loosening of monetary policy lowers short term interest rates
relative to long term rates, and to the extent that banks operate using short term
12 Theoretical models of such buffer stock holdings of bank capital include Passmore and Sharpe(1994), Baglioni and Cherubini (1994); Calem and Rob (1996); Froot and Stein (1998); Milne andWhalley (1999); Milne (2002); Van Den Heuvel (2002); Milne and Whalley (2003); Milne (2004);Van den Huevel (2004); Estrella (2004); Puera and Keppo (2006); and Zhu (2006) as a well as asmall empirical literature (see Jokipii and Milne (2008) for an empirical contribution and furthercitations).13 Kilponen and Milne (2008) analyse the interaction of banking sector book capital and optimal
monetary policy, finding that the resulting impact on output-inflation tradeoffs is small and onlyarises at all if bank loan rates have a cost channel impact on marginal costs of production.
-
7/29/2019 Bank Cycles 2009
40/62
38
funding for long term interest rate yielding assets, this increases bank net interest
income and hence eventually the book value of bank capital. At the same time the
cost of many bank liabilities are not sensitive to market rates of interest at all, so
lower market rates of interest can reduce bank net interest income and lowercapital. There is a similar ambiguity for the market value of bank capital. There
little direct impact from short term rates of interest, but market values of bank
capital depend critically on expectations of future bank earnings, of nominal and
real income growth, and of long term interest rates, and these may or may not
respond to shifts of monetary policy.
In the extreme, of which Japan in the early 1990s is be a good illustration,
monetary policy is unable to shift expectations of continued weakness of nominal
and real income growth. In such a case, even when monetary policy reduces short
term market interest rates to or close to zero, the demand for bank lending remains
low and capital constraints, partly through the low market value of bank assets,
continue to limit the supply of funds to banks, and thus the expectations of low
growth of incomes and of prices are confirmed.
Although the macroeconomic data for Japan described in Section 5 provide no
clear evidence on monetary transmission, their experience of monetary deflation
in the early 1990s supports our view that the bank balance sheets can attenuate
rather than amplify the impact of monetary policy. If the bank lending channel or
bank financial accelerator had been playing a quantitatively important amplifying
role, then the Bank of Japan would have found it much easier than they in fact did
to use monetary policy to stimulate deposit growth.Japan was able to begin its escape from monetary deflation only when, later in
the 1990s, Japanese banks finally properly recognised the scale of loan losses
occurred in the bubble economy of the late 1980s and were properly
recapitalised by the Japanese government. We conclude that constraints on bank
balance sheets not only limit the impact of monetary policy, but do so to a greater
extent when the banking sector is as a whole is severely undercapitalised.
Expansionary monetary policy is then not enough on its own to solve a systemic
under-capitalisation of the banking sector.
Such systemic undercapitalisation has once again recently emerged for banksin the US and Europe, with the high level of losses on sub-prime mortgage
securities and other structured credit products, and the resulting credit crunch in
which banks worldwide have been no longer able use mortgage backed securities
to raise low cost wholesale funding. It appears consistent with our analysis
that these funding problems have made it very difficult for banks to respond fully
to relaxation of monetary policy. The situation therefore requires not just
expansionary monetary policy response, but also the recognition of bank losses
and a restoration of net worth through the issue of new capital, exactly the actions
that have been adopted by financial authorities world wide in October of 2008 in
response to the worsening global banking crisis.
-
7/29/2019 Bank Cycles 2009
41/62
39
References
Altunbas, Y Fazylow, O Molyneux, P (2002) Evidence on the bank lending
channel in Europe. Journal of Banking and Finance 26 (11), 20932110.
Angelini, A Kashyap, A Mojon, B eds. (2003) Monetary Policy
Transmission in the Euro Area. Cambridge University Press.
Bernanke, B S Blinder, A S (1988) Credit, Money, and Aggregate Demand.
American Economic Review, (Papers and Proceedings), 78(2), 435439.
Bernanke, B S Blinder, A S (1992) The Federal Funds Rate and the Channels
of Monetary Transmission. American Economic Review, 82(4), 901921.
Bernanke, B S Gertler, M (1995) Inside the Black Box: The Credit Channel
of Monetary Policy Transmission. Journal of Economic Perspectives, Fall
1995, 9(4), 2748.
Bernanke, B S Gertler, M Gilchrist, S (xxxx) The financial accelerator in a
quantitative business cycle framework. Chapter 21, of J B Taylor and M
Woodford, The Handbook of Macroeconomics, Elsevier North Holland,
13411393.
Ehrmann, M Gambacorta, L Martinez-Pages, J Sevestre, P Worms, A
(2003) Financial systems and the role of banks in monetary policy
transmission in the Euro area. In I Angelini, A Kashyap, and B Mojon
(Eds.), Monetary Policy Transmission in the Euro Area, 33 Chapter 14, 235
269. Cambridge University Press.
Favero, C A Giavazzi, F Flabbi, L (2001) The transmission mechanism of
monetary policy in europe: Evidence from banks balance sheets.
Discussion Paper 2303,Centre for Economic Policy Research.
Freixas, X Rochet, J-C (1997) The Microconomics of Banking. The MIT
press: Cambridge, MA.
Friedman, B M Kuttner, K N (1993) Economic activity and short-term credit
markets: An analysis of prices and quantities. Brookings Papers on
Economic Activity (2), 193283.
-
7/29/2019 Bank Cycles 2009
42/62
40
Froot, K Stein, J (1998) A new approach to capital budgeting for financial
institutions. Journal of Financial Economics 47, 5582.
Gertler, M Gilchrist, S (1994) Monetary Policy, Business Cycles, and theBehavior of Small Manufacturing Firms. Quarterly Journal of Economics,
109(2), 309340.
Holmstrom, B Tirole, J (1997) Financial intermediation, loanable funds, and
the real sector. Quarterly Journal of Economics 112 (3), 663691.
Houston, J F James, C (1998) Do bank internal capital markets promote
lending? Journal of Banking and Finance 22, 899918.
Houston, J Christopher, J David, M (1997) Capital Market Frictions and the
Role of Internal Capital Markets in Banking. Journal of Financial
Economics, 46(2), 135164.
Hubbard, R G (1995) Is There a Credit Channel for Monetary Policy?
Federal Resewe Bank of St. Louis Review, 77(3), 6377.
Kashyap, A K Stean, J C (1994) Monetary Policy and Bank Lending. In N G
Mankiw, ed., Monetarypolicy. Chicago: University of Chicago Press, 221
256.
Kashyap, A K Stean, J C (1995) The Impact of Monetary Policy on Bank
Balance Sheets. Carnegie-Rochester Confererzce Series on Public Policy, 42,
151195.
Kashyap, A K Stein, J C (2000) What Do a Million Observations on Banks
Say about the Transmission of Monetary Policy? The American Economic
Review, Vol. 90, No. 3. 407428.
Kashyap, A K Stein, J C Wilsox, D W (1993) Monetary Policy and Credit
Conditions: Evidence from the Composition of External Finance.
American Economic Review, 83(1), 7898.
Kashyap, A K Stein, J C Wilsox, D W (1996) Monetary Policy and Credit
Conditions: Evidence from the Composition of External Finance: Reply.
American Economic Review, 86(1), 310314.
-
7/29/2019 Bank Cycles 2009
43/62
41
Kilponen, J Milne, A (2008) The Bank Lending Channel under Optimal
Choice of Monetary Policy. Bank of Finland Discussion Paper, 2008,
forthcoming.
King, S (2000) A credit channel in Europe: Evidence from banks balance
sheets. Mimeo, University of California: Davis.
Kishan, R P Opelia, T P (2000) Bank Size, Bank Capital and the Bank
Lending Channel. Journal of Money, Credit, and Banking, 32(1), 121141.
Kishan, R P Opelia, T P (2006) Bank capital and loan asymmetry in the
transmission of monetary policy. Journal of Banking and Finance, 30, 259
285.
Klein, M (1971) A Theory of the Banking Firm. Journal of Money, Credit, and
Banking 3, 205218.
Monti, M (1972) Deposit, credit, and interest rate determination under
alternative bank objectives. In Mathematical Methods in investment and
finance, edited by G P Szego and K Shell, Amsterdam: North-Holland.
Oliner, S D Rudebusch, G D (1996) Monetary policy and credit conditions:
Evidence from the composition of external finance: A comment. AmericanEconomic Review 86, 300309.
Peek, J Rosengren, E (1995) The Capital Crunch: Neither a Borrower nor a
Lender Be. Journal of Money, Credit, and Banking, 27(3), 625638.
Peek, J Rosengren, E (1997) The International Transmission of Financial
Shocks: The Case of Japan. American Economic Review, 87(4), 495505.
Peek, J Rosengreen, E S (2000) Collateral damage: Effects of the Japanesebank crisis on real activity in the united states. American Economic
Review 90(1), 3045.
Poole, W (1970) Optimal Choice of Monetary Policy Instruments in a Simple
Stochastic Macro Model. The Quarterly Journal of Economics, Vol. 84, No.
2 (May) 197216.
Romer, C D Romer, D H (1990) New evidence on the monetary transmission
mechanism. Brookings Papers on Economic Activity (1), 149198.
-
7/29/2019 Bank Cycles 2009
44/62
42
Stein, J C (1998) An Adverse Selection Model of Bank Asset and Liability
Management with Implications for the Transmission of Monetary Policy.
Rand Journal of Economics, 29(3), 466486.
Topi, J (2003) Effects of moral hazard and monitoring on monetary policy
transmission. Study E:24, Bank of Finland.
Woodford, M (2003) Interest and Prices: Foundations of a Theory of
Monetary Policy. Princeton University Press.
-
7/29/2019 Bank Cycles 2009
45/62
43
Appendix 1
Empirical studies of bank lending and monetary policy
Many studies examine the bank lending channel and the more general issue of the
role of bank credit in monetary transmission. Bernanke and Gertler (1995) review
the role of credit in monetary policy transmission, making the helpful distinction
between the balance sheet channel (also sometimes referred to as the financial
accelerator) and the bank lending channel. The balance sheet channel arises when
changes in the net worth of bank- dependent borrowers leads to an increase in
their cost of raising external finance. Most often this is interpreted as an increase
in bank monitoring costs, a mechanism which appears in several theoretical
models including Holmstrom and Tirole (1997).
The bank lending channel focuses not on borrowers, but on the effect of credit
market imperfections on the intermediation function of banks. Kashyap and Stein
(1994, 1995) state the following conditions for the presence of a bank lending
channel: (a) that some borrowers are bank dependent and cannot easily substitute
other forms of finance for bank lending; and (b) bank loan supply is affected by
bank balance sheet characteristics. Such supply impacts might arise through the
conventional bank lending channel mechanism but they might equally occur
because of constraints on net worth (equity capital) . Topi (2003) usefully
contrasts the conventional bank lending channel, as modeled by Bernanke andBlinder (1988) and Stein (1998), with the impact of capital constraints on the
supply of bank loans, arguing that the impact of capital constraints is likely to be
much more important. Other theoretical models of bank intermediation emphasise
the role of bank capital on loan supply, but do not model the transmission
mechanism of monetary policy.
There is a large body of evidence consistent with the presence of a balance
sheet channel. An influential study is Gertler and Gilchrist (1994) who find that
bank lending to small manufacturing firms varies much more than does bank
lending to larger firms, with changes in the net worth of firms. A large number ofother studies (reviewed by Bernanke and Gertler, 1995) which report cash flow
impacts on inventory and fixed capital investment also support the presence of a
balance sheet channel. The balance sheet mechanism is now routinely introduced
into dynamic general equilibrium models of monetary policy (beginning with
Bernanke, Gertler, and Gilchrist, 1999).
There is also a large empirical literature using similar specifications to the
studies of corporate balance sheet effects suggesting that capital, capital
regulation, and bank profits all affect the supply of bank lending. These include
Peek and Rosengreen (1993) who examine the role of capital regulations in